diff --git a/tensorflow_text/core/kernels/fast_wordpiece_tokenizer_model_builder.cc b/tensorflow_text/core/kernels/fast_wordpiece_tokenizer_model_builder.cc
index 9467c4d6e..f57056f63 100644
--- a/tensorflow_text/core/kernels/fast_wordpiece_tokenizer_model_builder.cc
+++ b/tensorflow_text/core/kernels/fast_wordpiece_tokenizer_model_builder.cc
@@ -217,7 +217,7 @@ class FastWordpieceBuilder {
   int unk_token_id_ = -1;
 
   // A wrapper to access the trie encoded by `trie_array_`.
-  absl::optional<trie_utils::DartsCloneTrieWrapper> trie_;
+  std::optional<trie_utils::DartsCloneTrieWrapper> trie_;
 
   // The actual data of the trie.
   std::vector<uint32_t> trie_array_;
@@ -271,7 +271,7 @@ absl::Status FastWordpieceBuilder::BuildModel(
   }
 
   // Determine `unk_token_id_`.
-  const absl::optional<int> unk_token_id = vocab_->LookupId(unk_token_);
+  const std::optional<int> unk_token_id = vocab_->LookupId(unk_token_);
   if (!unk_token_id.has_value()) {
     return absl::FailedPreconditionError("Cannot find unk_token in the vocab!");
   }
@@ -309,7 +309,7 @@ FastWordpieceBuilder::PrepareVocabTokensToBuildTrie() {
   std::vector<TrieVocabToken> tokens_to_build_trie;
   tokens_to_build_trie.reserve(vocab_->Size());
   for (int token_id = 0; token_id < vocab_->Size(); ++token_id) {
-    const absl::optional<absl::string_view> word = vocab_->LookupWord(token_id);
+    const std::optional<absl::string_view> word = vocab_->LookupWord(token_id);
     if (!word.has_value()) {
       return absl::FailedPreconditionError(
           "Impossible. `token_id` is definitely within the range of vocab "
@@ -837,7 +837,7 @@ absl::Status FastWordpieceBuilder::PrecomputeResultForSuffixIndicator() {
     return absl::FailedPreconditionError(status.error_msg);
   }
   for (int i = 0; i < subwords.size(); ++i) {
-    const absl::optional<int> subword_id = vocab_->LookupId(subwords[i]);
+    const std::optional<int> subword_id = vocab_->LookupId(subwords[i]);
     if (!subword_id.has_value()) {
       return absl::FailedPreconditionError(
           "Impossible because `subwords[i]` must be in the vocabulary!");
@@ -877,7 +877,7 @@ absl::StatusOr<std::string> FastWordpieceBuilder::ExportToFlatBuffer() const {
   if (support_detokenization_) {
     vocab_fbs_vector.reserve(vocab_->Size());
     for (int i = 0; i < vocab_->Size(); ++i) {
-      const absl::optional<absl::string_view> word = vocab_->LookupWord(i);
+      const std::optional<absl::string_view> word = vocab_->LookupWord(i);
       if (!word.has_value()) {
         return absl::FailedPreconditionError(
             "Impossible. `token_id` is definitely within the range of vocab "
diff --git a/tensorflow_text/core/kernels/phrase_tokenizer_model_builder.cc b/tensorflow_text/core/kernels/phrase_tokenizer_model_builder.cc
index 268aeb32c..cd464b467 100644
--- a/tensorflow_text/core/kernels/phrase_tokenizer_model_builder.cc
+++ b/tensorflow_text/core/kernels/phrase_tokenizer_model_builder.cc
@@ -71,7 +71,7 @@ absl::Status PhraseBuilder::BuildModel(const std::vector<std::string>& vocab,
   }
 
   // Determine `unk_token_id_`.
-  const absl::optional<int> unk_token_id = vocab_->LookupId(unk_token_);
+  const std::optional<int> unk_token_id = vocab_->LookupId(unk_token_);
   if (!unk_token_id.has_value()) {
     return absl::FailedPreconditionError("Cannot find unk_token in the vocab!");
   }
@@ -93,7 +93,7 @@ absl::StatusOr<std::string> PhraseBuilder::ExportToFlatBuffer() const {
   if (support_detokenization_) {
     vocab_fbs_vector.reserve(vocab_->Size());
     for (int i = 0; i < vocab_->Size(); ++i) {
-      const absl::optional<absl::string_view> word = vocab_->LookupWord(i);
+      const std::optional<absl::string_view> word = vocab_->LookupWord(i);
       if (!word.has_value()) {
         return absl::FailedPreconditionError(
             "Impossible. `token_id` is definitely within the range of vocab "
diff --git a/tensorflow_text/core/kernels/string_vocab.cc b/tensorflow_text/core/kernels/string_vocab.cc
index a2c239a93..509c0af81 100644
--- a/tensorflow_text/core/kernels/string_vocab.cc
+++ b/tensorflow_text/core/kernels/string_vocab.cc
@@ -30,7 +30,7 @@ LookupStatus StringVocab::Contains(absl::string_view key, bool* value) const {
   return LookupStatus();
 }
 
-absl::optional<int> StringVocab::LookupId(absl::string_view key) const {
+std::optional<int> StringVocab::LookupId(absl::string_view key) const {
   auto it = index_map_.find(key);
   if (it == index_map_.end()) {
     return absl::nullopt;
@@ -40,7 +40,7 @@ absl::optional<int> StringVocab::LookupId(absl::string_view key) const {
 }
 
 // Returns the key of `vocab_id` or empty if `vocab_id` is not valid.
-absl::optional<absl::string_view> StringVocab::LookupWord(int vocab_id) const {
+std::optional<absl::string_view> StringVocab::LookupWord(int vocab_id) const {
   if (vocab_id >= vocab_.size() || vocab_id < 0) {
     return absl::nullopt;
   }
diff --git a/tensorflow_text/core/kernels/string_vocab.h b/tensorflow_text/core/kernels/string_vocab.h
index 4590f2775..1bd481797 100644
--- a/tensorflow_text/core/kernels/string_vocab.h
+++ b/tensorflow_text/core/kernels/string_vocab.h
@@ -33,9 +33,9 @@ class StringVocab : public WordpieceVocab {
   StringVocab(const StringVocab&) = delete;
   StringVocab& operator=(const StringVocab&) = delete;
   LookupStatus Contains(absl::string_view key, bool* value) const override;
-  absl::optional<int> LookupId(absl::string_view key) const;
+  std::optional<int> LookupId(absl::string_view key) const;
   // Returns the key of `vocab_id` or empty if `vocab_id` is not valid.
-  absl::optional<absl::string_view> LookupWord(int vocab_id) const;
+  std::optional<absl::string_view> LookupWord(int vocab_id) const;
   int Size() const { return index_map_.size(); }
 
  private: